Hello,
I am analyzing 2 parameters which are called adiponectin and 1h-insulin. There is an inverse relationship between both. Low adiponectin levels are associated with high 1h-insulin levels. How can I determine the AUC and cut-off of adiponectin with such inverse relationship to 1h-insulin [hyperinsulinemic > 100 mU/mL is coded as 1) in a ROC curve in SPSS?
Thanks for the help.
Best regards, Moritz
Inverse regression is also called calibration. The probabilities are different. Google calibration regression to get a start on this. Best, David Booth
Thanks for the answer David. However I don't want to calculate the probability in a regression. I just want to create a ROC curve. So in SPSS, when I add adiponectin as test variable and 1h-insulin as state variable (value of state variable 1), I am getting a ROC which is curved in the opposite direction because of the inverse relationship between the 2 parameter. So I have to set value of state variable to "0" to receive a normal ROC curve. However I am not sure if this is the right approach? Thanks again, Moritz
Where are you going to get the probability values for the ROC curve?
Can I not just use the values of Adiponectin? Do I need the predicted probability?
https://blog.revolutionanalytics.com/2016/08/roc-curves-in-two-lines-of-code.html
The above gives an ROC curve for a binary classifier. See the link. As I understand it the prediction scores are prediction probabilities from the binary classifier i.e. outcomes from a logistic regression if that is the classifier. Am I missing something or did I say something wrong? Best, David
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS